PDF toolbox for DeepSeek Harness

Key features

• pdf_read toolExtract text page by page, with page number markers.
• Page number selection"1-3,5" or "all" - Large documents can be read in chunks.
• Document metadata (title) and total page count.
• Built-in boundary controlupper limit of file bytes, upper limit of pages parsed in a single time, upper limit of characters in a single call - the truncation behavior is clear and prompts the model how to continue.
• Read through the harness file system seam (ctx.fs), and the deployed permissions and sandbox policies automatically take effect.